Results of treatment of advanced-stage lymphoblastic lymphoma at St Jude Children's Research Hospital from 1962 to 2002

Abstract

Background

Reliable prognostic factors have not been established for advanced-stage pediatric lymphoblastic lymphoma (LL). We analyzed treatment outcomes and potential risk factors in children and adolescents with advanced-stage LL treated over a 40-year period.

Patients and methods

From 1962 through 2002, 146 patients (99 boys and 47 girls) with stage III (n = 111) or stage IV (n = 35) LL were treated at St Jude Children's Research Hospital. The five treatment eras were 1962–1975 (no protocol), 1975–1979 (NHL-75), 1979–1984 (Total 10 High), 1985–1992 (Pediatric Oncology Group protocol), and 1992–2002 (NHL13). Age at diagnosis was <10 years in 65 patients and ≥10 years in 81.

Results

Outcomes improved markedly over successive treatment eras. NHL13 produced the highest 5-year event-free survival (EFS) estimate (82.9% ± 6.1% [SE]) compared with only 20.0% ± 8.0% during the earliest era. Treatment era (P < 0.0001) and age at diagnosis (<10 years versus ≥10 years, P = 0.0153) were independent prognostic factors, whereas disease stage, lactate dehydrogenase level, and presence of a pleural effusion were not.

Conclusions

Treatment era and age were the most important prognostic factors for children with advanced-stage LL. We suggest that a better assessment of early treatment response may help to identify patients with drug-resistant disease who require more intensive therapy.

introduction

Approximately 30% of pediatric non-Hodgkin's lymphomas (NHL) are lymphoblastic lymphomas (LLs) [1, 2]. Modern treatment approaches for advanced-stage pediatric LL are generally derived from successful therapies for acute lymphoblastic leukemia (ALL). This strategy was initially validated by a randomized Children's Cancer Study Group (CCG) trial, demonstrating that the LSA₂L₂ regimen for ALL was superior to the alkylator-based COMP regimen for treatment of advanced-stage LL [3]. At least 75% of children with advanced-stage LL are long-term event-free survivors when treated with contemporary treatment regimens [3–6]. However, reliable prognostic factors for risk stratification to avoid over- or under-treatment have not been established for this patient population. Therefore, we sought to identify clinical features that are prognostically significant among 146 children with advanced-stage LL treated at our institution over a period of 4 decades.

patients and methods

patients and treatment

From 1962 to 2002, 146 children with advanced-stage LL were evaluated and treated at St. Jude Children's Research Hospital. Disease stage was defined as previously described [7]. From 1962 through 1975, patients were treated with various multi-agent combinations, including some designed for the treatment of ALL. From 1975 through 1979, all stages of disease were treated on an NHL-specific protocol with cyclophosphamide, (Baxter, Deerfield, IL) vincristine (Hospira, Lake Forest, IL), Adriamycin, prednisone, 6-mercaptopurine, and low-dose methotrexate [8]. From 1979 through 1984, children with advanced-stage disease were treated on the Total 10 High protocol which was designed for high-risk ALL and consisted of an antimetabolite backbone plus teniposide and cytarabine pulses [9]. Advanced-stage disease was treated from 1985 through 1992 with Pediatric Oncology Group protocol therapy [10]. From 1992 to 2002, patients with advanced-stage disease were treated on NHL13 [4], an institutional protocol similar to the total 13 regimen for T-cell ALL [11].

statistical analysis

Presenting features at diagnosis were analyzed using the Chi-square test. The principal end point was event-free survival (EFS), measured from the start of therapy to the date of first treatment failure (relapse, death, or second malignancy) or to the last date of follow-up. Failure to enter remission was considered an event at time zero. Distribution of EFS was estimated by the method of Kaplan and Meier and compared by using the log-rank test, stratified by treatment era where appropriate. Multivariable Cox proportional hazards regression models were used to assess whether EFS was associated with age at diagnosis, race, sex, mediastinal involvement, bone involvement, skin involvement, pleural effusion, CNS involvement, Murphy stage, or serum LDH activity.

results

presenting features

We studied the clinical course of 99 boys and 47 girls treated at our institution for stage III (n = 111) or stage IV (n = 35) LL. Table 1 summarizes patients' age, CNS involvement, serum LDH activity, and other presenting features. The prevalence of patient subgroups defined by these parameters did not differ significantly among the treatment eras (Table 1).

Table 1.

Selected presenting clinical and biological characteristics of 146 children with advanced-stage lymphoblastic lymphoma according to treatment era

Characteristic	Total (n = 146)	Treatment era					P value*
		PRE75 (n = 20) 1962–1975	NHL75 (n = 22) 1975–1979	TOTXH (n = 24) 1979–1984	POG (n = 39) 1985–1992	NHL13 (n = 41) 1992–2002
Age at diagnosis
Median (years)	11.4	8.4	9.6	12.0	12.6	12.1	0.35a
Age at diagnosis (%)
<10 years	65 (44.5)	13 (65.0)	12 (54.5)	9 (37.5)	14 (35.9)	17 (41.5)	0.19b
≥10 years	81 (55.5)	7 (35.0)	10 (45.5)	15 (62.5)	25 (64.1)	24 (58.5)
Sex (%)
Male	99 (67.8)	14 (70.0)	19 (86.4)	13 (54.2)	26 (66.7)	27 (65.9)	0.23b
Female	47 (32.2)	6 (30.0)	3 (13.6)	11 (45.8)	13 (33.3)	14 (34.1)
Mediastinal mass (%)
Yes	122 (83.6)	17 (85.0)	18 (81.8)	21 (87.5)	33 (84.6)	33 (80.5)	0.96b
No	24 (16.4)	3 (15.0)	4 (18.2)	3 (12.5)	6 (15.4)	8 (19.5)
Bone involvement (%)
Yes	15 (10.3)	1 (5.0)	2 (9.1)	1 (4.2)	6 (15.4)	5 (12.2)	0.58b
No	131 (89.7)	19 (95.0)	20 (90.9)	23 (95.8)	33 (84.6)	36 (87.8)
Marrow involvement (%)
Yes	24 (16.4)	2 (10.0)	3 (13.6)	8 (33.3)	5 (12.8)	6 (14.6)	0.19b
No	122 (83.6)	18 (90.0)	19 (86.4)	16 (66.7)	34 (87.2)	35 (85.4)
Skin involvement [n (%)]
Yes	2 (1.4)	0 (0)	0 (0)	0 (0)	0 (0)	2 (4.9)	0.27b
No	144 (98.6)	20 (100.0)	22 (100.0)	24 (100.0)	39 (100.0)	39 (95.1)
Pleural effusion [n (%)]
Yes	79 (54.1)	13 (65.0)	11 (50.0)	12 (50.0)	24 (61.5)	19 (46.3)	0.54b
No	67 (45.9)	7 (35.0)	11 (50.0)	12 (50.0)	15 (38.5)	22 (53.7)
CNS involvement [n (%)]
Yes	11 (7.5)	1 (5.0)	1 (3.8)	0 (0)	5 (12.8)	4 (9.8)	0.37b
No	135 (92.5)	19 (95.0)	21 (95.5)	24 (100.0)	32 (87.2)	37 (90.2)
Murphy stage [n (%)]
III	111 (76.0)	17 (85.0)	18 (81.8)	16 (66.7)	29 (74.4)	31 (75.6)	0.64b
IV	35 (24.0)	3 (15.0)	4 (18.2)	8 (33.3)	10 (25.6)	10 (24.4)
LDH activity
Median (U/l)	333	146	265	285	374	343	0.51a
<500 U/l [n (%)]	99 (67.8)	4 (20.0)	19 (86.4)	19 (79.2)	27 (69.2)	30 (73.2)	0.43b
≥500 U/l [n (%)]	31 (21.2)	1 (3.7)	3 (11.5)	4 (16.7)	12 (30.8)	11 (26.8)
Missing [n (%)]	16 (11.0)	15 (75.0)	0 (0)	1 (4.2)	0 (0)	0 (0)

^aWilcoxon rank-sum test.

^bPearson's Chi-square test or exact test for R×C table when necessary.

*P values reflect comparison across treatment eras, and ‘missing’ are excluded.

clinical outcome

Treatment outcomes differed markedly according to the treatment approach. For example, the 5-year EFS estimate was 20.0% ± 8.0% (SE) for the 20 patients treated before 1975, when no uniform therapy had been established; the EFS estimate increased to 66.7% ± 9.3% for the 24 patients treated between 1979 and 1984 and then to 82.9% ± 6.1% for the 41 patients enrolled in the most recent trial, NHL13 (P < 0.0001; Figure 1) [4]. Adverse events are summarized in Table 2.

Figure 1.

Event-free survival (EFS) estimates for treatment eras at St Jude: pre-NHL-75 (1962–1975), NHL-75 (1975–1979), Total X High (TOTXH, 1979–1984), and NHL-13 (1992–2002).

Table 2.

Summary of adverse events

Event	No NHL protocol–NHL75 (1962–1979) (n = 42)	TOTXH–POG (1979–1992) (n = 63)	NHL13 (1992–2002) (n = 41)	Total
Relapse
Hematologic	8	6	0	14
Hematologic + CNS	2	0	1	3
Local	5	9	2	16
CNS	2	2	0	4
Induction failure	7	3	2	12
Second malignancy	3	2	1	6
Toxicity-associated early death	7	1	1	9
Total	34	23	7	64

NHL, non-Hodgkin's lymphoma; TOTXH, total 10 high.

predictors of outcome

In addition to treatment era, age at diagnosis was the only feature significantly associated with the outcome. Patients <10 years of age had a higher EFS estimate than did older patients (5-year EFS, 64.6% ± 5.9% versus 55.5% ± 5.6%; P = 0.014) (Table 3). Both characteristics retained their independent prognostic significance (P < 0.0001 and P = 0.016, respectively) in the multivariable Cox proportional hazards regression model. When we examined the association of age (<10 years versus ≥10 years) with EFS within the three greater treatment eras, 1962–1979 (pre-1975 and NHL75), 1979–1992 (St. Jude Total 10H and POG), and 1992–2002 (NHL13), the age category remained significant during the total 10 high–POG era (1979–1992), but there was less power to detect a significant difference within the pre-NHL75–NHL75 era (1975–1979) or the NHL13 era (1992–2002), which had smaller patient numbers. In this regard, three of the five patients who experienced treatment failure on the NHL13 protocol were >10 years of age at diagnosis.

Table 3.

Event-free survival (EFS) according to presenting features and treatment era

		EFS estimate ± SE (%)
	n	Year 0	Year 5	Year 10	P valuea
Age at diagnosis (years)
<10	65	89.2 ± 3.6	64.6 ± 5.9	64.6 ± 6.5	0.014
≥10	81	93.8 ± 2.6	55.5 ± 5.7	55.5 ± 6.5
Sex
Male	99	89.9 ± 3.0	56.6 ± 5.0	56.6 ± 5.8	0.31
Female	47	97.9 ± 2.1	66.0 ± 7.3	66.0 ± 7.5
Mediastinal mass
Yes	122	91.0 ± 2.5	58.2 ± 4.6	58.2 ± 5.1	0.56
No	24	95.8 ± 4.0	66.7 ± 9.6	66.7 ± 11.1
Bone involvement
Yes	15	93.3 ± 6.2	80.0 ± 10.3	80.0 ± 11.3	0.26
No	131	91.6 ± 2.3	57.2 ± 4.4	57.2 ± 5.0
Skin involvement
Yes	2	100 ± 0.0	100 ± 0.0	100 ± 0.0	0.53
No	144	91.7 ± 2.2	59.0 ± 4.2	59.0 ± 4.7
Pleural effusion
Yes	79	88.6 ± 3.4	54.4 ± 5.7	54.4 ± 6.3	0.57
No	67	92.5 ± 3.1	65.5 ± 6.0	65.5 ± 7.0
Murphy stage
III	111	91.0 ± 2.6	61.2 ± 4.7	61.2 ± 4.9	0.44
IV	35	94.3 ± 3.8	54.3 ± 8.4	54.3 ± 8.9
LDH activity (U/l)
<500	99	96.0 ± 1.9	64.6 ± 4.9	64.6 ± 5.5	0.79†
≥500	31	90.3 ± 5.0	64.5 ± 8.8	64.5 ± 9.6
Data missing	16	68.8 ± 9.6	18.8 ± 8.5	18.8 ± 8.5
Treatment era
No NHL protocol (1962–1975)	20	75.0 ± 8.4	20.0 ± 8.0	20.0 ± 8.0	<0.0001
NHL75 (1975–1979)	22	90.9 ± 5.8	36.4 ± 9.7	36.4 ± 9.7
TOTXH (1979–1984)	24	95.8 ± 4.0	66.7 ± 9.3	66.7 ± 9.3
POG (1985–1992)	39	94.9 ± 3.4	64.1 ± 7.5	64.1 ± 7.8
NHL13 (1992–2002)	41	95.1 ± 3.3	82.9 ± 6.3	82.9 ± 8.8

^aLog-rank test. ^†LDH < 500 U/l versus ≥ 500 U/l.

discussion

The clinical outcome for pediatric LL improved progressively over the sequential treatment eras at our institution. Although improved supportive care contributed to this progress (as shown by seven toxicity-associated early deaths during 1962–1979 but only two during later eras; Table 2), more effective lymphoma treatment was most likely the determinant factor. To this end, the first significant advance, implemented in the total 10 high regimen, was the incorporation of teniposide and cytarabine into an otherwise standard antimetabolite-based ALL regimen [9]. The subsequent NHL13 regimen [4], derived from our total 13 protocol for high-risk ALL [12], added high-dose methotrexate in both consolidation and continuation phases, a re-induction phase (comprising additional L-asparaginase), and weekly administration of rotating drug pairs during continuation therapy. NHL13 yielded a 5-year EFS estimate of 83%. Importantly, this regimen did not include prophylactic cranial irradiation; CNS prophylaxis was achieved through systemic chemotherapy (high-dose methotrexate, intravenous etoposide, and oral dexamethasone) and extended administration of intrathecal methotrexate, hydrocortisone, and cytarabine during continuation. Thus, as in ALL, outcome was improved by intensification of chemotherapy and refinement of CNS prophylaxis. NHL13 was not randomized to definitively compare the contribution of individual features and many components are likely to have contributed to the excellent results. First, the addition of a re-induction phase with additional doses of L-asparaginase was probably important; in a randomized trial, Amylon et al [10] demonstrated that additional L-asparaginase improved the outcome of children with LL. Second, the incorporation of dexamethasone pulses throughout continuation is also likely to have been beneficial, as in childhood ALL [13]. Third, the use of etoposide during both the induction and continuation phases may have improved outcome, as suggested by this drug's known anti-ALL activity and its capacity to penetrate the blood-brain barrier [14]. Although the use of high-dose methotrexate in both consolidation and continuation (every 8 weeks for the first year) may have improved outcome in NHL13, there remains some controversy regarding the role of high-dose methotrexate in the treatment of advanced stage LL. In this regard, high-dose methotrexate during consolidation offered no benefit over extended intrathecal chemotherapy in a Children's Oncology Group LL study [15]. Moreover, a randomized trial carried out by the Pediatric Oncology Group demonstrated no outcome advantage for those children with advanced stage LL who received high-dose methotrexate [16].

We sought to identify presenting prognostic features. After adjustment for treatment era, only patients' age at diagnosis significantly related to outcome, indicating a worse prognosis for patients ≥10 years of age at diagnosis. However, within the three large treatment eras 1962–1979 (pre-1975 and NHL75), 1979–1992 (St. Jude Total 10H and POG), and 1992–2002 (NHL13), age group was significantly prognostic only during the 1979–1992 era, which had the largest sample size. It is possible that intensive contemporary ALL-based therapy has eliminated the prognostic significance of age. A multivariable analysis of patients treated on the BFM-90 protocol also failed to identify any of the clinical features that had prognostic significance [5]. Therefore, there are currently no reliable prognostic factors for patients with advanced LL.

Early response to therapy is the strongest independent predictor of treatment outcome in children with ALL [17–21]. The recent COG A5971 trial for LL used various diagnostic imaging modalities, including CT, gallium, and PET scanning, at several time points during induction therapy to determine whether early tumor response is related to treatment outcome [15]. However, diagnostic imaging studies have historically been unreliable in distinguishing necrotic from active residual masses [5]. PET imaging may improve our ability to detect residual disease, although confirmatory data for children with LL are lacking. The COG A5971 study also tested the prognostic significance of disease dissemination at diagnosis by using the same flow cytometric methods that are used to monitor minimal residual disease in children with T-cell ALL [17, 21–23]. The extent of minimal disseminated disease at diagnosis was found to be prognostically significant; patients with ≥1% circulating lymphoblasts had a significantly lower probability of EFS than did patients with a lesser extent of disseminated disease [24]. Based on these findings, our next study will use minimal disease measurements at diagnosis and during therapy for risk stratification.

funding

This work is supported in part by grant CA 21765 from the National Institutes of Health, by a Center of Excellence grant from the State of Tennessee, and by the American Lebanese Syrian Associated Charities (ALSAC).

disclosure

The authors have declared no conflicts of interest.